Exercise Training Rapidly Increases Hepatic Insulin Extraction in NAFLD

Abstract

Purpose:

We aimed to determine the immediacy of exercise intervention on liver-specific metabolic processes in Non-alcoholic fatty liver disease (NAFLD).

Methods:

We undertook a short-term (7-day) exercise training study (60 min/day treadmill walking at 80–85% of maximum heart rate) in obese adults (N=13, 58±3 years, 34.3±1.1 kg/m², >5% hepatic lipid by ¹H-MR spectroscopy). Insulin sensitivity (ISI) was estimated by oral glucose tolerance test using the Soonthorpun model. Hepatic insulin extraction (HIE) was calculated as the molar difference in AUC for insulin and C-peptide (HIE= 1-(AUC_Insulin÷AUC_C-Pep)).

Results:

The increase in HIE, VO_2max and ISI following the intervention was 9.8%, 9.8% and 34%, respectively (all P<0.05). Basal fat oxidation increased (Pre: 47±6 Vs Post: 65±6 mg/min, P<0.05) and carbohydrate oxidation decreased (Pre: 160±20 Vs Post: 112±15 mg/min, P<0.05) with exercise training. After the intervention, HIE correlated positively with adiponectin (r=0.56, P<0.05) and negatively with TNF-α (r=−0.78, P<0.001).

Conclusions:

By increasing hepatic insulin extraction along with peripheral insulin sensitivity, aerobic exercise training rapidly reverses some of the underlying physiological mechanisms associated with NAFLD, in a weight-loss independent manner. This reversal could potentially act through adipokine-related pathways.

Introduction:

Non-alcoholic fatty liver disease (NAFLD), characterized by excessive accumulation of fat in the liver, is commonly associated with obesity-related comorbidities such as type 2 diabetes, dyslipidemia and metabolic syndrome. It is estimated that the incidence of NAFLD in the US is approximately 25% of the adult population (1). The natural history of NAFLD includes a slow progression from simple steatosis (hepatocyte ballooning) to steatohepatitis (steatosis + inflammation) and then a faster progression towards fibrosis, cirrhosis and in a more limited number of cases, hepatocellular carcinoma (2). The severity of NAFLD and its progression towards cirrhosis and liver failure is associated with escalating risk towards atherogenesis and cardiovascular disease (3,4). Slowing down, or possibly reversing, the progression of NAFLD to advanced fibrosis has an immense potential to reduce the societal cardiovascular and general health risk burden.

There are no effective and safe pharmacological agents for treating NAFLD (5), as metformin and thiazolidinediones have been shown to improve liver histology only after inflammation sets in (commonly referred to as non-alcoholic steatohepatitis or NASH). Previously, it was reported that bariatric surgery induced long-term improvements in liver histology in patients with grade-3 obesity and NASH (6,7). Interestingly, the long-term effectiveness of bariatric surgery for improving liver pathology was predicted by early improvements in whole-body insulin sensitivity. Weight loss (~5–10%) achieved by a combination of physical activity and hypocaloric diet is the mainstay of therapy for NAFLD and NASH (5). However, the amount weight loss achieved, even in a tightly regulated setting, is highly variable (8). In this study, we sought to determine whether short-term aerobic exercise training could manifest metabolic improvement in liver related outcomes.

The liver plays a critical role in maintaining peripheral insulin concentrations relative to insulin secretion by regulating the first-pass metabolism of insulin, a phenomenon known as ‘hepatic insulin extraction’ (HIE). In healthy adults, ~50–80% of the secreted insulin (in the portal vein) is extracted and cleared in the hepatocytes (9). Meier et al. found that both the amount of insulin secreted in a pulsatile fashion and the amplitude of the pulsatile peak were both positively correlated with hepatic insulin clearance and delivery of insulin to the circulation (9). Incretins, like GLP-1, can also theoretically influence HIE primarily by modulating the degree of insulin secretion. While the presence of hepatic insulin resistance is established in NAFLD (10), the relationship with liver fat content is strongly debated (11,12). Understanding these intricate relationships is important in order to devise targeted therapies to manage NAFLD. In this study, we explored whether HIE is related to liver fat content and/or peripheral insulin resistance. Secondly, we evaluated whether HIE is modulated by aerobic exercise, independent of weight loss.

Methods

The recruitment and methods have been previously described (13). Briefly, as shown schematically in Fig 1, thirteen obese, sedentary adults (age 58±3.4 years; BMI 34.3±1.1 kg/m²) with radiographically-confirmed NAFLD (>5% intra-hepatic lipid content) were enrolled into a short-term, 7-day exercise intervention. After obtaining informed consent, individuals underwent a medical screening, physical examination, oral glucose tolerance test (OGTT), and blood analyses including a lipid panel and liver enzymes to estimate hepatic function. The exclusion criteria included individuals with any active disease condition; taking any medications known to affect study outcomes; individual alcohol consumption >20 g/d for men and >10 g/d for women; involved in any on-going exercise or weight loss regime; any contraindications to exercise (as detected during a 12-lead EKG exercise test); and post-menopausal status or hormonal replacement therapy in females. This study was approved by Cleveland Clinic Institutional Review Board. All participants signed informed consent forms prior to study participation.

Figure 1. Schematic of Study Design.

Procedures listed according to the chronological sequence of the day. See Methods for more details.

Cardiorespiratory Fitness:

Physical activity levels were estimated using the Minnesota Leisure Time Physical Activity Questionnaire (14). The subjects were deemed sedentary if they expended <300 kcal/day of energy through their leisure time activity. An incremental graded treadmill exercise test was conducted to determine the participant’s maximal oxygen consumption (VO_2max) as previously reported (15). Expired air was continuously sampled using an automated sampler (Jaeger Oxycon Pro, Viasys, Yorba Linda CA). Heart rate was continuously captured during each exercise session. Maximal heart rate was used to prescribe exercise intensity during the exercise training sessions (Polar Electro Inc., Woodbury, NY). The test was deemed satisfactory if ≥3 of the following criteria were attained: (1) a respiratory quotient of >1.10; (2) self-determined fatigue; (3) heart rate at <10 beats per min of age-predicted maximum; (4) plateau in oxygen consumption with increasing workloads. Participants abstained from coffee and alcohol consumption for 12 and 48 hours, respectively prior to exercise testing.

Body Composition:

Height and body weight were measured using standard techniques. Whole body fat distribution was measured by dual-energy X-ray absorptiometry (Lunar model iDXA; Madison, WI).

Aerobic Exercise Intervention:

All of the subjects participated in 60 minutes of supervised aerobic exercise on a treadmill at ~85% of HR_max each day for seven consecutive days. They were instructed to maintain current diet during this intervention period.

Substrate Metabolism:

After an overnight fast, subjects were admitted to the Clinical Research Unit, rested for ~30 mins and indirect calorimetry was conducted to estimate energy expenditure and substrate metabolism. Exhaled air was continuously sampled for ~15 minutes using an automated system under a ventilated hood until steady state was achieved. Energy expenditure, and rates of carbohydrate and fat oxidation were calculated as previously described (16).

Glucose Metabolism:

Following indirect calorimetry, a standard 75-g OGTT was performed at ~8:00 am before and after the 7-day exercise intervention. An intravenous catheter was placed in an antecubital vein and secured with a saline lock. After the baseline draws, the glucose drink was consumed and whole blood samples were drawn in plastic tubes containing EDTA at 30, 60, 90, 120 and 180 minutes after ingestion. Incremental area under the curve (iAUC) was calculated using the trapezoidal equation after baseline correction. Insulin sensitivity Index (ISI) was estimated using the Soonthorpun model (17):

$$\text{ISI}\left(\text{OGTT}\right)=\frac{\left[\frac{1.9}{6}*\text{BW X FPG}+520-\frac{1.9}{18}*\text{BW}*\text{AUCglu}-\frac{\text{Uglu}}{1.8}\right]*1000}{\left(\text{AUCins\hspace{0.17em}}*\text{\hspace{0.17em}BW}\right)}$$

Here, BW = body weight in kg, AUC_glu = area under the glucose curve following glucose load for 3 hours, U_glu = urinary loss of glucose, and AUC_ins = area under the insulin curve for 3 hours during the OGTT. Hepatic insulin extraction (HIE) was calculated from molar areas under the insulin and C-peptide curves and is expressed as fold change relative to baseline (18). Hepatic insulin resistance index (HIRI) was estimated as the product of AUC_gluc and AUC_ins during the first 30 minutes of the test (19). The C-peptide-genic index, which is the increase in the peripheral C-peptide concentration relative to increase in glucose concentration within 30 minutes of oral glucose loading, was calculated from [ΔC-peptide_(30–0)÷ ΔGlucose_(30–0)] (19).

Plasma Analyses:

Plasma glucose was determined at the time of the test using a YSI 2300 STAT Plus (Yellow Springs, OH). The rest of the plasma was immediately stored at −80⁰C. Plasma insulin concentration was determined using a radioimmunoassay (Millipore, Billerica, MA). Plasma C-peptide was measured using ELISA (Linco Research, St. Charles, MO). Plasma free fatty acid concentrations (FFA) were determined using a colorimetric assay (Wako Pure Chemical Industries, Richmond VA). GLP-1 and HMW adiponectin were measured through the 120-min time point of the OGTT using commercially available enzyme-linked immunosorbent assay (ELISA) kits (Millipore, St. Charles, MO). MNC-derived TNF-α was determined via high-sensitivity ELISA (R & D Systems, Minneapolis, MN).

Liver Scans:

Intrahepatic lipid content (IHL) was measured by ¹H-magnetic resonance spectroscopy (MRS) on a 3T MR system (Siemens Sonata, Erlangen, Germany), before and after the exercise intervention. Following an overnight fast, the subjects arrived at the laboratory at ~6:00 am. A body array MRI coil, with center aligning to the subject’s spine and shoulders, was affixed using Velcro straps. The subjects were placed in a prone position and head first into the scanner on a memory foam mattress to further minimize respiratory artifact. To accurately scan the liver, an 8-cm³ voxel was placed within the right posterior lobe of the subject’s liver, and MR spectra with and without water suppression were acquired with a single-voxel PRESS acquisition (TR=5,000 ms and TE=30ms) (20). Final data were Fourier-transformed, filtered, baseline-corrected and phased. The diagnosis of NAFLD was confirmed if IHL was >5%.

Statistical Analyses:

Normality of the data was determined using the Shapiro-Wilk test. Continuous variables are expressed as mean and standard error of the mean for parametric measures, and median and inter-quartile ranges for non-parametric measures. Categorical variables are expressed as percentages. Differences between the pre-intervention and post-intervention groups were analyzed for statistical significance using paired Student t-tests (parametric), and Wilcoxon-signed rank test (non-parametric) for continuous variables and Chi-squared test for categorical variables; α was assumed at 0.05. Effect size was calculated using Cohen’s D formula as the difference in the means divided by the standard deviation of the difference. With the current paired sample size of 13, effect size of 0.84, standard deviation of 0.075, α of 5%, the calculated power for our primary outcome of HIE was >95%. Correlation analyses were performed using Pearson’s and Spearman methods and the association between biochemical and anthropometric parameters were evaluated using multivariate linear regression, with and without subgroup analysis, after power transformation. All graphical and descriptive statistical analyses were performed using RStudio - version 1.1.453.

Results:

Data from a subset of the individuals included herein have been previously reported (13,21); this secondary analysis is focused on HIE and hepatic insulin resistance. Participant characteristics are shown in Table 1. Blood and metabolic screening data revealed that subjects were pre-diabetic with a fasting blood glucose concentration of 5.9±0.3 mM and met the diagnostic criteria for hepatic steatosis based on IHL measured with MRS (AVG: 14%±9%).

Table 1.

Subject Characteristics. iAUC: integrated Area Under the Curve; FPG: Fasting Plasma Glucose; Data presented as mean±SE.

Parameters	Pre- Study	Post- Study	P
Age (years)	57.5±3.6	---	---
Weight (kg)	97±4	96±4	0.63
Body Fat (%)	44.9±1.7	44.6±1.6	0.19
VO2max (mL/kg/min)	23.6±1.7	25.7±1.7	0.005
FPG (mM)	5.9±0.3	5.7±0.2	0.66
iAUC Glucose (180 min)	66.9±16.3	52.9±14.5	0.002
Fasting Plasma Insulin (pM)	149.1±18.1	126.4±12.1	0.04
iAUC Insulin (180 min)	111908±14841.2	80166.7±10407.7	0.005
Fasting C-peptide (nM)	962.1±126.5	832.0±91.4	0.08
iAUC C-Peptide (180 min)	258510±30971	229320±30186	0.41
Fasting GLP-1 (pg/mL)	11.7±2.6	10.1±2.8	0.03
iAUC GLP-1 (180 min)	1832.3±369.2	1539.9±313.3	0.004
Adiponectin (ng/mL)	3513±653	3659±767	0.58
TNF-α (pg/L)	1963±241	2639±500	0.58
Liver Triglyceride Content (%)	14.0±2.9	14.1±2.2	0.7
Hepatic Polyunsaturated Fat Fraction (%)	0.4±0.08	0.5±0.07	0.02
Carbohydrate Oxidation Rate (g/min)	0.16±0.02	0.11±0.02	0.006
Lipid Oxidation Rate (g/min)	0.05±0.007	0.07±0.005	0.07

The 7-day aerobic exercise training prescription induced a 9.8±3.1% increase in VO_2max (P<0.01), and 20.7±6.5% and 21.5±10.4% decrease in the iAUC for glucose and insulin after the glucose load (P<0.05), respectively (Fig 2). There was no change in body weight or body fat percentage after the exercise program. Though fasting glucose concentration and C-peptide did not change with exercise training, peripheral insulin concentrations decreased significantly (11.9%±4.3%, P<0.05). There was also an 11.5±4.2% and 13.5±3.9% (P<0.05) decrease in fasting and iAUC for GLP-1 after the exercise intervention, respectively (Fig 2). Other plasma metabolic and biochemical parameters did not significantly change.

Figure 2.

Profiles and areas under the curve for glucose, insulin, C-Peptide and GLP-1 during the OGTT. a.u.: arbitrary units. *Post Vs Pre-intervention, P<0.05

HIE and ISI_OGTT increased significantly (9.8±3.2% and 49.2±14.7%, respectively, P<0.05). HIRI decreased by 4.0±18% from baseline (P<0.05) (Fig 3). The C-peptide-genic index trended toward decrease, however this was not statistically significant (Pre:721.6±222.5 Vs Post: 489.6±90.5 pM/mM, P=0.17). There was a 63.5±32.9% increase in hepatic polyunsaturated lipid content (P<0.05), though the saturated fat percentage and the triglyceride content did not differ for these participants. Whole-body fasting carbohydrate oxidation rate decreased by 28.2±8.1% (P<0.01).

Figure 3.

Differences in OGTT-derived indices of HIE, HIRI and ISI, before and after the exercise intervention. *Post Vs pre-intervention, P<0.05

Correlation Analyses:

At baseline, a higher HIE percentage correlated with a lower HIRI, higher peripheral insulin sensitivity, lower hepatic saturated fat percentage and higher plasma HDL concentration (r=−0.5, 0.67, −0.48 and 0.6, respectively, all P<0.05) (Fig 4). The change in HIE correlated significantly with the change in peripheral insulin sensitivity and the change in peak glucose concentration after OGTT (r= −0.69 and 0.61, respectively; P<0.05) (Fig 5). After the intervention, HIE correlated positively with adiponectin (r=0.56, P<0.05) and negatively with plasma TNF-α concentration (r= −0.78, P<0.001).

Figure 4.

Relationship between baseline HIE and - A) HIRI; B) ISI; C) hepatic saturated fat (%) and D) plasma HDL concentration at baseline. Solid line represents regression line, while the dashed lines represents the 95% confidence intervals.

Figure 5.

Link between the change in HIE and change in peak glucose concentration at 90 minutes post-exercise intervention.

Discussion

NAFLD is a complex disease that arises from a combination of reduced insulin sensitivity (both hepatic and peripheral), hyperinsulinemia and their related downstream metabolic effects. There is conflicting evidence that HIE is associated with liver fat content, owing to differences in methodologies used and study populations (11,18,22). In this study, we found that lower HIE was significantly associated with a higher saturated fraction of liver fat. The positive association of plasma HDL with HIE was further increased by its interaction with body weight (adj-R²= 0.3 Vs 0.58, P<0.05), potentially through ApoA-1 (23), which suggests a central role for insulin dynamics in the development and the progression of NAFLD. ApoA-1 is the major protein component of HDL and is found to have an inverse relationship with hepatic fat content. Decreased HIE, and thus, prevailing hyperinsulinemia promotes insulin resistance (24,25) (Fig 4). More recently, Braun et al. described elevations in plasma insulin concentrations after a 7-day increase in sedentary time owing to dynamic reductions in total HIE (AUC) (26). Hence, exploiting a therapeutic option which can independently affect each of these components in a dose-dependent and time-dependent manner, should seemingly result in a robust treatment response.

The American Association for the Study of Liver Diseases treatment guidelines recommend exercise as the most important component of treatment for NAFLD (5). The physiologic adaptations to exercise are multifactorial, however, to date, the independent effects of exercise on NAFLD have not been isolated from concomitant weight loss. Our study provides evidence that aerobic exercise training quickly reverses some of the underlying pathophysiological derangements of NAFLD, even before the weight-loss phase sets in. These effects may be due to discrete and interdependent increments in hepatic insulin extraction, and peripheral and hepatic insulin sensitivity (27). The associations between changes in HIE and peripheral and hepatic insulin resistance support our working hypothesis. Given that skeletal muscle is the biggest insulin responsive organ in the body, it stands to reason that the change in skeletal muscle insulin sensitivity may be contributing to the change in HIE. Although the short-term exercise intervention used in this study did not significantly reduce liver triglyceride content or change percentage of saturated fat, small but likely clinically important changes were predicted by the interaction between changes in HIE and VO_2max. Our findings are in agreement with Utzschneider et al., who compared HIE in NAFLD and non-NAFLD subjects and determined that HIE was not related to liver fat but instead was related to peripheral insulin resistance (11).

Exercise is known to independently change both glucose and insulin kinetics (28). Fasting plasma glucose and C-peptide concentrations did not change after the intervention. However, the insulin concentrations required to maintain blood glucose levels decreased significantly, indicating the important contribution of increased HIE to maintaining overall glucose homeostasis. In support, we observed that the overall change in HIE correlated positively with change in the peak glucose concentration (90 min after OGTT) (Fig-5 change in HIE and glucose), a measure that also contributes significantly to reduced CVD risk (29).

As previously published (30), there was a decrease in fasting and total GLP-1 content in plasma with short-term exercise training in these patients. However, we did not observe any significant statistical associations between changes in HIE and GLP-1 nor any significant change in insulin secretion based on the C-peptide-genic index. This suggests that exercise-induced improvements in glucose homeostasis in people with NAFLD are potentially due to early increase in HIE and are independent of modulations of the GLP-1/insulin-secretion pathway. The distinct changes in GLP-1 may eventually contribute to sustained improvements in the capacity to secrete insulin. Furthermore, the significant relationship between HIE and adiponectin (r=0.6) and TNF-α (r=−0.78) after exercise intervention highlights the contribution of HIE to a metabolically healthy and lower systemic inflammatory milieu. These effects may be mediated through circulating free fatty acid concentrations, which could directly reduce HIE independent of prevailing glucose concentrations (31).

First-pass extraction of insulin is ~50–80%, and is predominately driven by the liver (32), followed by the kidney at ~20% (33). In the liver, HIE involves binding and internalization via receptor-mediated endocytosis and later degradation by insulin-degrading enzyme (IDE) via lysosome proteolysis (34). Hyperinsulinemia has been implicated to reduce HIE by insulin-receptor saturation (35), though it was recently counter-argued that acute changes in HIE are essential for dampening systemic insulin oscillations relative to pulsatile insulin secreted from β-cells (9). Kirwan et al. (36), have previously shown that acute exercise increases HIE in untrained adults, which may be attributed to increased expression of hepatic IDE (37). Kurauti et al. (37) found that exercise can increase expression and activity of IDE at least via IL-6 (myokine) mediated- transduction and activation of STAT3 in liver. Long-term aerobic exercise training, increases mitochondrial biogenesis, activates hepatic PGC1-α expression, and alters hepatic metabolism by altering the IRS1:IRS2 ratio (38), eventually leading to decreased steatosis. Thus, chronic exercise training could potentially lead to systemic PPAR-α activation and reduced CEACAM1-mediated HIE in the context of improved disposition index (39). We speculate that the increase in HIE in our study may be due to early changes in PPAR-α and mitochondrial biogenesis. We also found that whole-body carbohydrate oxidation rate was decreased in our participants. It is still unclear if this finding is due to change in HIE or peripheral insulin sensitivity, or both. The exploration regarding the potential mechanistic links between substrate oxidation and HIE is still underway.

One of the potential limitations of our study is the use of OGTT modelling for estimating HIE, instead of the gold standard hepatic portal vein sampling. We minimized the error of model estimation during the non-steady state by using area-under-the-curve calculations of C-peptide and insulin after the concentration curves returned to baseline levels (40). Other potential limitations could be that our study population is not gender-matched, limited sample size, non-exercising controls, or absence of isotope tracers.

Conclusions

In conclusion, we found that the pathophysiology contributing to NAFLD is quickly reversed by 7 days of aerobic exercise training. Even in the absence of weight loss, exercise reduced hyperinsulinemia through an increase in hepatic insulin extraction, and concomitantly, muscle and liver insulin sensitivity. These effects were independent of changes in fasting and aggregate plasma incretin concentrations. Changes in total GLP-1 and adiponectin sensitivity may add more potency to the effectiveness of longer-term exercise in these patients. Our findings further strengthen the rationale for the use of exercise in managing NAFLD.